In the video surveillance system, basic functions such as real-time video browsing and real-time video recording and so on can be provided. However, if it is required to implement labeling and corresponding alarm for a target video, it is required to perform real-time analysis on the video.
Moreover, in some specific application scenarios, it requires pretty high accuracy and stability in video tracking, such as in human face tracking and recognition. Where, the accuracy refers to that the labeling location in video tracking is accurate, while the stability refers to that the labeling information is consecutive in video image displaying.
In the related art, there is no solution on how to analyze, track and label a specific video accurately.